Depends on the mics.
To take a couple of frequently-used examples:
First up, the Shure SM58. As a live sound engineer, I have to deal with these pretty often. They've been around almost forever, and musicians are comfortable seeing them. They're the minimum standard of any serious venue. They're far from my favourite mic, but that's another story. Typical usage is a rock band vocalist screaming into it at point-blank range.
Info here:
http://recordinghacks.com/microphones/Shure/SM58
Frequency response drops very quickly at 10kHz. Having measured a couple, I can say that it falls off a cliff there. You can't EQ it back. Given that, you could argue for a 20kHz sample rate, and you wouldn't miss much from one of the most popular mics on the planet.
Bit depth is more tricky. In theory, dynamic mics (mylar diaphragm with a coil attached. Coil moves in a magnetic field, generating voltage) have virtually unlimited maximum SPL, so a very high bit depth is called for. Microphones do get connected to a preamp, though, and those do have a limited dynamic range - at the low level, they have intrinsic noise relating to the Johnson Noise of the input resistors. At a high level, they clip. The required bit depth can thus be calculated.
At the opposite end of the microphone spectrum, we have the Schoeps MK4:
https://schoeps.de/en/products/colette/capsules/cardioids/mk-4.html
This is one of the mics that will get chosen for high-end classical recording, picking up an entire orchestra. With the latest CMC1 preamp attached, it's good for 135dBSPL at the high end, and has an Equivalent Noise Level of 14dB(A). ie, 121dB of dynamic range.
Unfortunately, Schoeps doesn't publish the frequency response in the ultrasonic region. I can say, though, that my MK41 (supercardioid version) can pick up ultrasonic pest repellers working, putting out 25kHz + 50kHz. The frequency response of those repellers is unknown, but the spectral analysis showed 50kHz being around 10dB down compared to 25kHz. I used a MOTU M4 running at 192kHz, feeding into REW's RTA.
Given that these mics can clearly capture something well into the ultrasonic region, a high sample rate could be called for, if one wishes to capture 100% of the mic's signal.
Finally, it's worth noting that virtually every recording you've ever heard will have some dynamic processing. Compressors, limiters, gates and expanders. A compressor might take a 24-bit microphone feed and reduce the dynamic range to 20-bit. A gate will increase the dynamic range infinitely, but aren't used where you might care. Limiters are a sub-set of compressors. Expanders do occasionally get used in classical mixing, and can increase the dynamic range of the mic feed beyond what was there originally. Used carefully, they can produce excellent results.
Even so, the final mix will be subject to compressors/limiting to get the ratio of quiet/loud sections "correct", which puts a final upper limit on the theoretical dynamic range of a piece of music.
The mixing desk that I use for multi-track recording (I can post some clips if anyone's interested in what real, uncompressed instruments sound like) is 48kHz/32-bit-float, and I've found it hasn't been a limiting factor on the quality of my recordings, even when I bring out the Schoeps mics.
Chris